A Highly Extensible Framework for Molecule Dynamic Simulation on GPUs
Xiao Zhang1, Wan Guo1, Xiao Qin2 and Xiaonan Zhao1
1School of Computer Science Northwestern Polytechnical University 127 Youyi xi Road, Xi’an Shaanxi China
2Department of Computer Science and Software Engineering Auburn University, AL 36849-5347
E-mail: zhangxiao@nwpu.edu.cn;xilouyouki@163.com;xqin@auburn.edu;zhaoxn@nwpu.edu.cn
Molecular dynamics (MD) was widely used in chemistry and bio molecules. Numerous attempts have been made to accelerate MD simulations. CUDA enabled NVIDIA Graphics processing units (GPUs) use as a general purpose parallel computer chips as CPU. But it is not easy to port a program to GPU. We present a highly extensible framework for molecular dynamics simulation. And we discuss how to accelerate the process of port to GPU. We introduce how to find the performance battle and how to port the time costly procedure to GPU. We discuss about how to decrease the memory usage in GPU and how to improve the maintenance of molecular dynamics simulation. At last, we present the performance of linear and parallel simulation with different number of molecules. Source codes can be found at https://github.com/orlandoacevedo/MCGPU.
Proc. the International Conference on Parallel and Distributed Processing Techniques and Applications, July 22-25, 2013
Acknowledgments: This research was supported by the U.S. National Science Foundation under Grants CCF-0845257 (CAREER), CNS-0917137 (CSR), CNS-0757778 (CSR), CCF-0742187 (CPA), CNS-0831502 (CyberTrust), CNS-0855251 (CRI), OCI-0753305 (CI-TEAM), DUE-0837341 (CCLI), and DUE-0830831 (SFS).